05:07 < amish> hii

05:07 < amish> how to start??

05:08 < amish> der??

13:05 < vlad_gl> hi! i posted comment with armadillo inplace_trans testing - https://github.com/mlpack/mlpack/issues/203. what do you think about this?

14:03 < awhitesong> naywhayare: Hi. I went through the source of superLU . I guess they are using dgssv.c for lu factorization and dgstrf.c for solving and not sure but i think there's no separate header for different types of sparse matrices now. and also they have supermtrix.h for their sparse matrix implementation.

14:08 < awhitesong> Also In Eigen, they have their own implementation of sparse matrices as class SparseMatrix<Type> in sparsecore/sparsematrix.h and have SuperLUSupport.h where these superLU headers are wrapped.

14:16 < awhitesong> So now if we have sparse matrices implemented in armadillo, then we can make a LUsolve<armadillo_sparse_type> class as an interface to another class having lu() and solve() functions wrapping ggssv.c and dgstrf.c from superLU. Please see this and if there's anything i can help then please tell me. Thanks.

19:00 < naywhayare> awhitesong1: thanks for taking a look into SuperLU. I'll find a little time on Monday and I'll let you know what I come up with. in the mean time, if you want to help with the Armadillo contribution, take a look at how ARPACK and LAPACK are wrapped in armadillo

19:01 < naywhayare> that's done in armadillo_bits/arpack_wrapper.hpp and armadillo_bits/arpack_bones.hpp (and lapack_bones.hpp and lapack_wrapper.hpp)

19:02 < naywhayare> I'm going to try to build an interface kind of like that for SuperLU, which can provide type-independent versions of the SuperLU functions

19:03 < naywhayare> then in sp_auxlib_bones.hpp/sp_auxlib_meat.hpp, I'll write an lu() and solve() function that use those type-independent SuperLU functions

19:03 < naywhayare> and then all that needs to be written is the rest of the glue code, like lu(SpMat&) in fn_lu.hpp which calls sp_auxlib::lu(), and so forth

19:04 < naywhayare> the armadillo code can be a little hard to understand inside, so if you don't think it'll be fun to implement, I'll take care of it (or at least parts of it) on Monday :)

19:06 < naywhayare> awhitesong: I sent some messages, but I dunno if maybe they went to the wrong client. you can see the IRC logs here, though: http://mlpack.org/irc/

19:08 < stephentu> naywhayare: is this superLU stuff to implement a sparse Ax=b solver?

19:10 < naywhayare> stephentu: yeah, which can then be used with ARPACK to get eigenvectors for a system Ax = Bx\lambda

19:12 < stephentu> cool stuff